Corresponding author: G. Scardia, Istituto di Geologia Ambientale e Geoingegneria, CNR, via Salaria km 29.300, I-00016 Monterotondo Scalo, Italy. (firstname.lastname@example.org)
 This interdisciplinary study on the subsurface stratigraphy of the Po Plain (northern Italy) brings new evidence in support of a climate-driven erosional unloading of the Alps since the Middle Pleistocene. A newly acquired, high-resolution seismic profile and a critical review of industrial seismic lines were integrated with sedimentologic observations on four magnetostratigraphically dated continental cores to reconstruct a three-sequence evolution of the Pleistocene clastic infill in the northern Po basin. During the first sequence (PS1; ∼1.4–0.87 Ma), characterized by sedimentation rates of ∼34 cm/kyr outpacing regional subsidence, meandering river systems prograded over the basin passing downstream to a cyclothemic shelfal succession. The second sequence (PS2; ∼0.87–0.45 Ma), heralded by a regional unconformity (R surface) correlated to the onset of the major Pleistocene glaciations, was characterized by widespread continental sedimentation of generally distal braidplain. The third sequence (PS3; ∼0.45 Ma to present), marked at the base by another regional unconformity (Y surface), is characterized by proximal braided fluvial deposition under combined conditions of confinement, erosion, and bypass. We interpret the PS3 sequence as deposited under the effects of a flexural uplift of the northern Po Plain during the Middle Pleistocene starting at ∼0.45 Ma, in response to the long-term erosional unloading of the Alps triggered by the waxing and waning of Alpine glaciers since the late Early Pleistocene global cooling (∼0.9 Ma). According to our modeling, erosion on a relatively limited area of the Alpine mountain chain, ranging from 1.3 to 1.7 mm/yr in the axial sector to 0.1–0.3 mm/yr at the margins, has been able to trigger rock uplift over a wider area including the proximal peripheral basins.
 The present study aims at improving our knowledge on the Pleistocene rock uplift of the Alps with an interdisciplinary study of the subsurface stratigraphy of the Po Plain, northern Italy (Figure 1). In recent years, the interpretation of selected industrial seismic profiles led to the recognition of regional reflectors in the Po basin, associated to important changes in style of sediment deposition and basin evolution [Di Dio, 1998; Carcano and Piccin, 2002; Ghielmi et al., 2010] (Appendix A). One of these reflectors, the R surface, has been related to the onset of major Pleistocene glaciations in the Alps during the late Early Pleistocene global cooling, culminated with marine isotope stage (MIS) 22 at ∼0.9 Ma [Muttoni et al., 2003; Elderfield et al., 2012]. At shallower depths above the R surface, another regional seismic reflector, here named Y surface, has been recognized [Di Dio, 1998; Carcano and Piccin, 2002]. In this study, both reflectors are accurately resolved by means of high-resolution reflection seismics, facies analysis of four continental cores drilled in the Milan area, and magnetochronology. The time-calibrated vertical and horizontal facies architecture is then integrated with previous studies in order to define the main depositional sequences of the uppermost ∼200 m of the Po Plain. We then explore an erosion-driven flexural model of the Alps to explain the middle Pleistocene uplift of the northern Po Plain and the formation of the Y surface, which we interpret it occurred in response of enhanced physical erosion triggered by the waxing and waning of Alpine glaciers since the late Early Pleistocene global cooling.
2. Regional Background
 The Po Plain is located between the Alps and the Apennines thrust belts and evolved as a foreland basin of the Apennines since the Messinian (Figure 1) [Pieri and Groppi, 1981]. Apennines-related flexural subsidence involved also the southern sector of the Alps, resulting in the southward tilting and burial of the outermost Alpine thrusts [e.g.,Fantoni et al., 2004]. During the Pliocene and the Early Pleistocene (now including the Gelasian stage [Gibbard et al., 2010]), the stepwise northward migration of the Apennine thrusts produced basinwide structural modifications and the deposition of a number of tectonostratigraphic sequences [Ghielmi et al., 2010]. Since the Middle Pleistocene, the outward migration of the Apennines stalled, and no new thrusts developed [Picotti and Pazzaglia, 2008; Ghielmi et al., 2010].
 The Milan area evolved as a distal ramp of the Apennines foreland basin during the Pliocene. The folded Miocene bedrock is sealed by three shelf sequences (LM, EP, and PL4; Figure 2) separated by two main regional unconformities produced, respectively, by the intra-Zanclean and Gelasian Apennine tectonic phases [Ghielmi et al., 2010]. Above these shelf sequences lies the PS1 stratigraphic sequence, biostratigraphically constrained in the Gaggiano 1 well to the Early Pleistocene (Figure 2) and consisting of mud and sand layers interpreted as basinal turbidites. In the late Early Pleistocene, a major progradation of the fluvial and deltaic systems occurred, probably in response to a reduction in subsidence rate; at this time, bathymetry decreased from neritic to shelfal (Trenno 1 and Gaggiano 1; Figure 2). The PS1 shelf is characterized by the coarsening-upward cyclothems observed in the resistivity logs of Trenno 1 and Gaggiano 1 (Figures 2 and 3) and produced by the glacioeustatic oscillations on shelfal settings [Scardia et al., 2006].
 Above PS1, Ghielmi et al.  identified the PS2 stratigraphic sequence marked at the base by the R surface, related to the onset of major Pleistocene glaciations in the Alps [Muttoni et al., 2003]. PS2 is often poorly resolved by industrial seismic profiles and represents the main target of this study by means of a high-resolution seismic line in the Milan area and four continental drill cores chronologically constrained with magnetostratigraphy.
3. High-Resolution Reflection Seismics
 A N–S trending, 710 m long, 96-channel seismic profile (hereafter BSC line) has been acquired at the west end of Milan [De Franco et al., 2009] (Figure 1). Details on data acquisition and processing are reported in Appendix B. The final stack of the BSC line was converted to depth using the results of the velocity analysis. In Figure 3, the mean interval velocities and the standard deviation band are displayed together with the velocity logs previously measured for Trenno 1 and Gaggiano 1.
 From ground surface, the BSC mean velocity increases linearly to ∼85 m. A velocity inversion layer with a minimum velocity value of 1.60 km/s was detected between 85 m and 160 m, followed by a second inversion layer with a minimum observed velocity of 1.76 km/s located between 196 m and 296 m. The two layers are bounded at the top by a velocity value of 1.72 km/s and 1.83 km/s, respectively. Below the second layer, the velocity curve increases quite regularly from ∼1.90 km/s at 330 m depth up to ∼2.40 km/s at 955 m depth. A comparison between the synthetic velocity pattern obtained from the BSC line acquisition and the measured velocity values of Trenno 1 and Gaggiano 1 shows a good agreement, considering the standard deviation band and the different resolution of the well data with respect to the BSC line.
 The conjunct use of high-resolution seismics, the localized waveform stack at Common Depth Points (CDP) 95–105, and the Trenno 1 resistivity log allows correlating high-amplitude reflectors with intervals of high-amplitude resistivity contrast observed in Trenno 1 at depths of ∼490–450 m, ∼310–170 m, and ∼100–35 m (Figure 3). As expected, the depositional environment structure yields a strong control by on the acoustic response, where seismic reflections are physical surfaces separating lithologies with strongly contrasting acoustic impedance (e.g., fines versus gravel/sand layers).
 By integrating the multidisciplinary information gathered by electric logs and regional seismic stratigraphy, we have recognized in the BSC line an overall regressive sequence characterized by four different seismic facies, described hereafter from bottom to top (Figure 3): (1) a pack of layered, high-amplitude, laterally discontinuous reflectors, characterized in the Trenno 1 resistivity log by coarsening upward cycles, which are interpreted as shelfal marine/continental cyclothems of glacioeustatic origin similar to those observed in the Po basin byAmorosi et al. , Kent et al. , and Scardia et al. , (2) a pack of low-amplitude, laterally discontinuous reflectors, characterized in the Trenno 1 resistivity log by fine-grained lithologies arranged in fining-upward cycles and attributed to low-energy fluvial environments, (3) a pack of layered high-amplitude, laterally continuous reflectors, interpreted as fluvial deposits of alternating high/low energy, and (4) an interval rich in seismic diffractions and characterized by poorly defined reflectors, probably related to the occurrence of massive, coarse-grained fluvial deposits.
 Facies 1 and 2 belong to the PS1 sequence, whereas the shift to facies 3 corresponds to the regional seismic horizon known as R surface referred to the onset of major Pleistocene glaciations and the southward progradation of glacial outwash deposits (Figures 2–4) [Muttoni et al., 2003]. Seismic facies 3 belongs to the PS2 sequence; its layered structure suggests a distal braidplain, made up mainly of alternating sand and gravel bodies attributed to glacial/interglacial cycles or arising from a fluvial depositional style characterized by repeated lateral migration of the channel network. The passage from seismic facies 3 to 4 corresponds to another regional seismic reflector, the Y surface (Figures 3 and 4), across which the well-layered seismic facies 3 passes upward into a chaotic assemblage of reflectors, interpreted as a sharp change of fluvial depositional style. As we shall see, the Y surface marks the onset of a new sequence, named here PS3 following the nomenclature ofGhielmi et al. . A similar vertical succession of seismic facies (PS1–PS3 sequences) has been recognized also elsewhere in the Po Plain, e.g., ∼15 km eastward of the BSC line [Francese et al., 2005].
4. Core Lithostratigraphy
 High-resolution seismic data of the uppermost ∼200 m were calibrated and interpreted by means of four continental cores drilled in the Milan area (Figure 1, inset). The southernmost cores (Peschiera Borromeo RL8 and Gaggiano RL9; Figures 5 and 6) reached a depth of 180 m and 148 m, respectively, whereas the northernmost cores (Milano Triulza RL10 and Milano Parco Nord RL11; Figures 6 and 7) reached the maximum depth of 100 m. A total of 528 m of sediments were recovered (Table 1). RL10 was drilled ∼4 km north of the Trenno 1 well and the BSC line (Figure 1, inset).
Latitude and longitude according to the WGS84 reference system; altitude in meters above sea level; total depth of cores in meters from the drill site ground surface; N, number of paleomagnetic samples collected; n, number of paleomagnetic samples used to outline the magnetic polarity stratigraphy.
Milano Parco Nord
 The overall lithostratigraphy consists of three superposed continental depositional systems, bounded by the regional seismic reflectors (unconformities) R and Y (Figure 4), and corresponding to the PS1, PS2, and PS3 stratigraphic sequences, as described hereafter from bottom to top.
4.1. Meandering River Plain
 This basal depositional system correlates to seismic facies 2 in the BSC line and belongs to sequence PS1. It shows a succession of unfossiliferous silt and clay packages alternated with fine- to medium-grained sand, often arranged in thin fining-upward cycles, each of them commonly attaining one to few meters in thickness. This facies association is well expressed in core RL8 (Figure 5) from core bottom (180 m) up to 121 m, where the presence of laminated silt and clay with intercalated sharp-based, fining-upward sand, and organic-rich layers points to floodplain deposits with dominant overbank deposition punctuated by crevasse splay episodes. In core RL9 (Figure 6), from core bottom (148 m) up to 109 m, massive silt and organic-rich layers are overlain by a major stack of amalgamated sand bodies (140–111 m), mainly consisting of horizontally laminated sand and interpreted as fluvial channel deposits. In core RL10 (Figure 7), the depositional system is represented from core bottom up to 73 m by discrete fining-upward cycles, consisting of well-sorted, massive or cross-laminated sand bodies, interpreted as fluvial channels passing upward to channel abandonment fines. On the whole, the association of overbank fines and fluvial channel sands points to a meandering river depositional system [e.g.,Miall, 2006].
4.2. Distal Braidplain
 This intermediate depositional system correlates to seismic facies 3 in the BSC line and belongs to sequence PS2. It starts with an abrupt boundary at its base, and consists of thick packages of alternating medium- to coarse-grained sand, gravel, and subordinate silt. Gravel layers are crudely bedded, whereas sand intervals are massive or cross laminated, often enriched in sparse pebbles. Gravel and sand are arranged in fining-upward, amalgamated (core RL9;Figure 6) or discrete (core RL10; Figure 7) cycles. Fine-grained deposits mostly consist of unfossiliferous, massive silt and clay, usually occurring at top of the fining-upward cycles.
 The overall facies association documents an organized fluvial system, where gravel and sand facies represent high- to medium-energy fluvial channels, with fines depositing in abandoned channels and floodplain settings. We interpret this depositional system as a distal braidplain with wandering fluvial channels. The abrupt change between the PS1 meandering river plain (section 4.1) and the PS2 distal braidplain corresponds to the R surface observed in the seismic profiles (Figures 2–4) and correlated to the onset of major Pleistocene glaciations in the Alps [Muttoni et al., 2003]; hence, the distal braidplain fluvial system is interpreted as the distal outwash plain produced by Alpine valley glaciers [e.g., Zielinski and Van Loon, 2003] during glacial stages.
4.3. Proximal Braidplain
 This upper depositional system belongs to the newly defined PS3 sequence and correlates to seismic facies 4 in the BSC line. Its base is tentatively placed in cores RL8 and RL9 at ∼48 m, and in cores RL10 and RL11 at ∼21 m and ∼42 m, respectively (Figures 5–8). This depth range is consistent with the depth of the Y reflector in the BSC line and in other studied cores in the Po Plain (Figure 4) [Scardia et al., 2006]. The Y reflector marks a strong increase of clast-supported gravel with respect to medium- to coarse-grained sand and pebbly sand. Crudely bedded gravel is the overwhelming facies with local thin intervals of pebbly sand, interpreted on the whole as incomplete or amalgamated fining-upward cycles. The almost complete absence of sand layers or well-developed fining-upward cycles suggests the occurrence of important erosional processes, likely due to a confined, unstable network of laterally shifting fluvial channels [see alsoOri, 1993].
 The observed facies association points to a high-energy fluvial system interpreted as a proximal braidplain, where the vertical and lateral stack of dominant coarse-grained fluvial deposits likely explains the scattered reflection observed in the seismic facies 4. Based on the available chronologic constrains provided by magnetostratigraphy [Scardia et al., 2010], these braidplain deposits are coeval to the Middle Pleistocene moraines and glaciofluvial deposits preserved a few tens of kilometers northward [e.g., Bini et al., 2004] (Figure 1, inset) and can therefore be referred to a proximal outwash plain produced by the Alpine valley glaciers.
 We applied paleomagnetism as a tool to overcome the traditional dating problems in continental settings. By measuring the magnetization of sediments as function of depth (i.e., time), it is possible to obtain a magnetic polarity stratigraphy insofar as magnetizations characterized by northerly (southerly) directions and positive (negative) inclinations are interpreted as acquired during normal (reverse) polarity intervals of the Earth's magnetic field of known age [e.g., Lourens et al., 2005].
 Paleomagnetic properties were studied on a total of 80 cubic (∼8 cm3) samples collected from cohesive, fine-grained sediments (Table 1). Sampling frequency was on the order of one sample every 7–8 m save for intervals dominated by coarse-grained sediments such as in core RL11 (Figure 8), where only a few samples could be collected. Details on laboratory analyses are provided in Appendix C.
 Isothermal remanent magnetization (IRM) acquisition curves of representative samples (Figure 9) show an initial steep growth at low fields with most of the magnetization acquired within ∼0.3 T fields, followed by a gentle slope with no tendency to saturate up to a maximum applied field of 2.5 T (Figure 9, samples RL8–30.3, RL8–132.4, and RL10–54.3). These results are compatible with the presence of magnetic minerals with contrasting coercivities interpreted as magnetite and hematite, as deduced also from the maximum unblocking temperatures of the medium (0.4 T) and high (2.5 T) coercivity components of, respectively, ∼570°C and ∼680°C (Figure 9); in some samples (e.g., RL10-74.3;Figure 9), hematite is the dominant magnetic phase. In association with magnetite and hematite, minor amounts of Fe-sulfides have been observed in a few samples.
 The intensity of the natural remanent magnetization (NRM) is on the order of 10−3–10−4 A/m with 20% of the samples in the 10−2 A/m range. Orthogonal projections of demagnetization data typically indicate the existence of a lower unblocking temperature component superimposed to a higher unblocking temperature component (Figure 10). The lower unblocking temperature component, removed between room temperature and a maximum of ∼250°C, usually bears steep positive (down pointing) inclinations, and is regarded as due to a (sub) recent magnetization overprint. The higher temperature (characteristic) component was removed to the origin of the demagnetization axes mainly in the magnetite and hematite temperature ranges between ∼350 and ∼680°C (see unblocking T in Figures 5–8). This characteristic component (ChRM) bears either positive (down pointing) or negative (up pointing) inclinations and is regarded as acquired at or shortly after sediment deposition as detrital remanent magnetization (DRM) or postdepositional DRM. Contrasting inclination values have been observed in samples from RL10 above 51 m depth. We suspect the occurrence in this interval of a normal polarity overprint that could not be fully removed during demagnetization treatment. This magnetic overprint is thought to have chemically originated as a consequence of iron mobilization produced by groundwater level oscillations.
 Cores were not oriented in azimuth during drilling, hence only the inclination of the ChRM component was used to delineate magnetic polarity stratigraphy. Statistical analysis of inclination-only data [Arason and Levi, 2010] revealed mean ChRM inclinations (Table 2) shallower than the geocentric axial dipole (GAD) inclination of ∼64° for the mean site latitude (∼N45°30′; Table 1), likely because of depositional inclination errors or postdepositional compaction of the sediments [e.g., Tauxe, 2005].
Table 2. Statistic Values of the Paleomagnetic Inclination Data
6. Magnetostratigraphy and Regional Correlation
 Magnetic polarity stratigraphy was interpreted by comparison with the ATNTS2004 time scale [Lourens et al., 2005] assuming as first-order approximation that depth is a linear function of time. We also took into account previous magnetostratigraphic and biostratigraphic data provided by several additional cores from the Po Plain (seeFigure 4 for location of all the cores) studied in the last decade (core RL2 is from Muttoni et al. ; cores RL1, RL3, RL4, RL5, RL7 are from Scardia et al. ). These data allowed us to generate a correlation framework valid for the uppermost ∼200 m of the proximal Po basin infill over a distance of ∼120 km from the westernmost core RL3 to the easternmost core RL1 and comprising the cores investigated in this study (Figure 11 and Table 3).
Table 3. Magnetochronology of the Uppermost ∼200 m of the Northern Po Basina
 On the whole, the upper 63 ± 13 m of the northern Po basin sediments have a normal magnetic polarity referred to the Brunhes chron (C1n; present–0.781 Ma). The underlying sediments down to 126 ± 20 m, bearing reverse magnetic polarity, deposited during the late Matuyama chron (C1r.1r; 0.781–0.988 Ma). Below, a normal polarity magnetozone follows down to the depth of 151 ± 21 m and is interpreted as the Jaramillo subchron (C1r.1n; 0.988–1.072 Ma). Finally, the lowermost reverse polarity identified in some cores down to a maximum depth of 220 m (i.e., cores RL2, RL3, RL4, RL7) is ascribed to the middle Matuyama chron (C1r.2r; 1.072–1.778 Ma). According to this chronostratigraphic framework, we recognized from top to bottom a virtually complete Brunhes–late Matuyama–Jaramillo succession in cores RL8 and RL9, whereas the poorly resolved reverse polarity magnetozone in cores RL10 and RL11 is interpreted as pertaining to a generic (late?) Matuyama subchron.
7. Sediment Accumulation Rates
 The chronologic constraints provided by the C1r.2r/1n (1.072 Ma), C1r.1n/1r (0.988 Ma), and C1r.1r/C1n (0.781 Ma) magnetic polarity reversals in core RL8 imply sediment accumulation rates of ∼41 cm/kyr during the late Early Pleistocene. In core RL9, sediment accumulation rates can be only roughly estimated in the range of ∼11–26 cm/kyr because of the uncertainties related to the C1r.1n/1r polarity reversal depth and the absence of the C1r.2r/1n reversal. The calculated rates are consistent with sediment accumulation rates of other cores reported by Scardia et al. , ranging from ∼24 cm/kyr (core RL7) to 42 cm/kyr (core RL1).
 Taking into account the whole data set, the mean sedimentation rate in the northern Po Plain during the late Early Pleistocene (∼1.5–0.78 Ma) is assessed at 34 ± 3.4 cm/kyr by regression analysis (Figure 12). The shallowing-upward trend displayed by the late Early Pleistocene shallow marine to continental cyclothems in cores RL1, RL4, RL5, and RL7 (Figure 11) implies that subsidence did not fully keep pace with sedimentation, which caused progressive infilling of the Po basin during the Jaramillo subchron (1.07–0.99 Ma) and the following late Matuyama subchron (0.99–0.78 Ma).
 Chronologic constraints above the Brunhes–Matuyama boundary (0.781 Ma) are very poor, thus hampering a direct evaluation of the sedimentation rates. As formerly discussed by Scardia et al. , the age of continental sediments deposited during the Brunhes chron can be modeled by adopting two end-member hypotheses. Assigning a “zero” age to the ground surface leads to a strong decrease in sedimentation rates to regional values of ∼10–6 cm/kyr during the Middle–Late Pleistocene. By instead extrapolating the average Early Pleistocene sediment accumulation rates to ground surface, an age range of ∼0.6–0.5 Ma is obtained for sediments exposed at or immediately below the ground surface. Both hypotheses suggest on the whole that the proximal Po basin experienced during the Middle–Late Pleistocene an overall decrease in sedimentation rates and/or erosional episodes.
8. Pleistocene Sedimentation in the Northern Po Basin
 The lower PS1 sequence deposited during the late Early Pleistocene (Matuyama chron; ∼1.4–0.87 Ma) and consists of meandering river plain deposits, sourced mainly in the Western and Central Alps [Muttoni et al., 2003; Garzanti et al., 2011] and prograding axially in low subsidence settings. Because of the progressive infilling trend of the Po basin from west to the east, the meandering river system passed eastward (toward Venice) to cyclic alternations of shallow marine (shelf and prodelta) and transitional deposits (delta plain, beach, and lagoon). These cyclothemic patterns are produced by climatic-driven processes and related to glacioeustatic oscillations [Amorosi et al., 1999; Kent et al., 2002; Scardia et al., 2006]. For instance, during a sea level fall and lowstand, transitional to continental coarsening-upward and/or shoaling-upward sequences developed, sealed by marine deposits during the following sea level rise. As a whole, PS1 represents a shallowing-upward sequence rapidly filling ∼300 m of residual accommodation space (∼0.3 two-way time seconds from base to top of clinoforms;Figure 2). The overall river progradation from west to the east makes the vertical transition from marine to continental deposits diachronic at the basin scale inasmuch as it becomes progressively younger eastward. As a consequence, in the western cores (RL3, RL8, RL11, RL10, RL9), cyclothems lie below the cored succession as documented in the BSC line (seismic facies 1) and in the Trenno 1 well (Figure 3).
 The onset of Pleistocene major glaciations in the Alps occurred within the late Matuyama and produced the R surface at ∼0.87 Ma [Muttoni et al., 2003], which is a major sequence boundary in the Po basin (Figure 4) marking the substantially synchronous and widespread progradation of PS2 braidplain deposits over the PS1 meandering river plain (Figure 11). PS2 straddles the Brunhes–Matuyama boundary (0.781 Ma) and consists of braidplain deposits, attributed to the distal fringe of the glacial outwash plains, prograding transversally to the basin axis from the Central and Southern Alps southward [Muttoni et al., 2003; Garzanti et al., 2011].
 The boundary with the uppermost PS3 is marked by the regional Y surface (Figure 4). According to magnetostratigraphic constraints, the Y surface falls systematically in the Brunhes chron (<0.781 Ma). Garzanti et al.  tentatively ascribed the Y surface to MIS 16 lowstand at ∼0.63 Ma by counting (undated) glacial/interglacial cycles starting from the R surface at MIS 22 [Muttoni et al., 2003]. By tracing the reflector from the Po basin to the Eni wells in the Adriatic Sea, the Y surface seems to occur just above the Last Occurrence of age diagnostic nannofossil Pseudoemiliania lacunosa [Carcano and Piccin, 2002] at 0.44 Ma [Lourens et al., 2005]. This datum is used here to infer an age of ∼0.45 Ma for the Y surface, albeit a precise age remains rather poorly determined. PS3 deposited during the Middle–Late Pleistocene and consists of proximal braidplain deposits prograding over the PS2 distal braidplain deposits (Figure 11). In detail, PS3 is composed almost exclusively of coarse-grained, poorly sorted gravels, interpreted as a stack of proximal outwash plain deposits, bounded by low-rank erosional surfaces. This fluvial system increases its internal organization downstream toward the more sagging sectors of the Po basin, where also the fine-grained interglacial deposits are preserved (e.g., south of Lake Garda [Amorosi et al., 2008]).
9. The Middle Pleistocene Flexural Uplift of the Alps
 We observe that the onset of rock uplift in the northern Po Plain [Scardia et al., 2006] and the shift in fluvial style between PS2 and PS3 occurred roughly at the same time, i.e., during the Middle Pleistocene. A number of causes have been proposed to explain the observed rock uplift in the greater Alpine region including deep-seated processes of uncertain magnitude and time scale (see discussion byChampagnac et al. ). Particularly appealing in its robust simplicity is the hypothesis of Heller et al. that the erosion-driven flexural rebound of an orogenic belt, e.g., the Alps, can trigger uplift and erosion in the proximal sector of the peripheral basins, e.g., the northern Po Plain, where long-term uplift rates on the order of ∼0.1 mm/yr with respect to the sea level have been observed since the Middle Pleistocene [Scardia et al., 2006] and are comparable to modern rates inferred from geodetic measurements [Arca and Beretta, 1985; Schlatter et al., 2005], attributed to erosion [Champagnac et al., 2009].
 Considering the lack of active convergence in this sector of the Alps [Delacou et al., 2004] since at least the Pliocene [Fantoni et al., 2004; Picotti and Pazzaglia, 2008; Ghielmi et al., 2010], we support the hypothesis that a common driving mechanism is still acting over the proximal basin since the Middle Pleistocene. Moving from this consideration, we tested the flexural response of the northern Po Plain to the erosional unloading of the Alps by fitting present-day geodetic data from a 20 km wide swath trending from Basel across the Alps to Milan (Figures 13 and 14a) with a simple 2-D model [Jordan, 1981]. These data are obtained by repeated precise leveling measures of a same benchmark over a time interval ranging from years to many decades. Arca and Beretta  expressed their measurements relative to sea level at Genoa (Italy), while data from Schlatter et al.  are relative to Aarburg (Switzerland), assumed to be stable (0 mm/yr). As exhumation at any benchmark is zero, the observed surface uplift must totally equate the rock uplift at the measurement spot except for subsiding areas, where other common processes such as sediment compaction and load may bias the tectonic signal [England and Molnar, 1990].
 By assuming an initial state of isostatic equilibrium, geodetic uplift rates can be directly converted into length units of vertical deflection from the initial equilibrium state [Champagnac et al., 2009]. The elastic component of uplift was studied by modeling the flexural uplift of an infinite beam composed of 20 km wide blocks. As erosion is expected to act mainly in the Alpine relief, only those blocks located from 70 to 230 km (see horizontal axis in Figure 14b) were modeled using a nonlinear least squares MATLAB solver and assuming crustal density (ρc) of 2700 kg/m3, mantle density (ρm) of 3300 kg/m3, Young's modulus (E) of 70 GPa, and Poisson's ratio (ν)) of 0.25 [Watts, 2001; Turcotte and Schubert, 2002]. For each block, the actual elastic thickness of the (Adriatic or European) lithosphere (Te) and the unloading (upthrust) value were determined simultaneously by inversion analysis starting from initial ranges of probable values of Te (ranging from 15 to 25 km [e.g., Royden, 1993; Allen et al., 2001]) and unloading (ranging from 0 mm (no upthrust) to −5 mm (maximum upthrust)). The best fit solution gave a Te value of 19.9 km, corresponding to a flexural rigidity of 4.97 × 1022 N m, and the unloading distribution reported in Figure 14b. The mean residual of the solution (Figure 14b) is 0.152 mm, and it is comparable with the mean error value of the fitted geodetic data (0.110 mm). The estimated unloading values due to erosion are on the order of 1.3–1.7 mm in the axial sector of the Alps and 0.1–0.3 mm at the margins, in fair agreement with denudation values estimated with other methods [e.g., Hinderer, 2001; Wittmann et al., 2007; Delunel et al., 2010; Norton et al., 2010]. It has been shown that denudation rates are spatially correlated with both elevation and present-day rock uplift [e.g.,Wittmann et al., 2007] and probably coupled [Champagnac et al., 2009]. Our model supports the isostatic flexural rebound of the Alps as the main driving mechanism to explain the coupling between erosion and rock uplift, acting over a 105year time through both long-term regional erosion produced by glacial waxing and waning cycles [e.g.,Champagnac et al., 2007] and short-term local-scale denudation processes such as frost cracking [Delunel et al., 2010] and slope dynamics [Norton et al., 2010].
 If the erosional unloading were fully Airy compensated (i.e., Te = 0), the ratio between uplift and erosion would be homogenous all over the mountain belt. However, as the lithosphere strength is involved in the isostatic rebound (in this case Te = ∼20 km), the uplift/erosion ratio changes across the mountain belt, ranging in our calculation from a minimum of ∼0.5 in the area of maximum erosion (i.e., the axial chain) to maxima of ∼3.0 and ∼1.2 in the northern (Swiss) and southern (Po Plain) proximal basins, respectively, where erosion is low. The average value weighted by area is 0.76, in agreement with the theoretic value of ∼0.8 determined by the crust versus mantle density contrast, expected if rock uplift was entirely driven by erosional unloading [Molnar and England, 1990].
 In summary, our model shows that, because of the flexural rigidity of the lithosphere, erosion affecting a limited area (i.e., the mountain chain) is able to trigger rock uplift over a wider area, spreading out off the belt to the peripheral basins with uplift/erosion ratios even larger than in the mountain chain [e.g., Cederbom et al., 2004; Scardia et al., 2006; Champagnac et al., 2008]. The estimated rebound wavelength is ∼270 km, calculated between the flexural nodes (points where deflection is zero) at both sides of the orogen. Interestingly, subsidence outward from the flexural nodes in both distal foreland basins (Swiss, Po Plain) is theoretically expected. Even if small, increasing accommodation space due to flexural subsidence can trigger sediment accumulation and loading (not modeled in our calculation) in a positive feedback mechanism.
10. The Y Surface
 The Y surface marks a regional change in fluvial depositional style, from distal braidplain with an organized network of wandering channels (PS2) to proximal braidplain (PS3) characterized by a confined network of laterally shifting fluvial channels, likely experiencing frequent episodes of cannibalization. Abrupt changes in fluvial deposition style can be attributed to various causes such as migration of the watersheds or exposure of more erodible rocks. Petrographic and fission tracks studies on Pleistocene sediments from the Po Plain, however, do not support geological or geomorphologic changes in source areas testifying for substantial watershed stability (see discussion by Garzanti et al. ). Actually, episodic watershed changes in the Plio-Pleistocene Alps seem to be confined in tectonically active areas, such as the Eastern Alps [e.g.,Monegato and Stefani, 2010].
 Basing on the evidence brought by our modeling, we interpret the Y surface as marking the transition from a subsidence-dominated stage to an uplift-dominated stage occurring in the proximal Po basin at ∼0.45 Ma in response of the flexural rebound of the Alps. The related rock uplift transformed the northern Po Plain from a depositional area into a limited accommodation space area characterized by dominant processes of sediment bypass. Enhanced climate-driven erosion is documented in the Alpine axial chain at ∼0.8 Ma [Häuselmann et al., 2007] roughly since the onset of the major Pleistocene glaciations (∼0.87 Ma [Muttoni et al., 2003]), but the flexural unloading is recorded in the proximal Po basin only in the Middle Pleistocene, probably since ∼0.45 Ma. Also taking into account the uncertainties associated to the dating of the Y surface, mainly based on biostratigraphic data, a lag time of ∼0.4 Myr between onset of glacial cycles and uplift is apparent. Such a lag time can be probably ascribed to a complex, and not yet fully resolved, interaction between the waxing erosional-driven uplift of the Alps and the waning tectonic subsidence related to Apennines northward migration [Picotti and Pazzaglia, 2008; Ghielmi et al., 2010].
 Our interpretation differs substantially from Garzanti et al. , who tentatively ascribed the Y surface to MIS 16 lowstand at ∼0.63 Ma by means of simple MIS counting starting from the R surface at MIS 22. In our view, the Y surface does not simply (or only) marks a glacial lowstand, but rather it signals a profound climate-driven tectonic control on the depositional style of the proximal Po basin.
 A peripheral basin of the Alps, the Po Plain, evolved during the Pleistocene in response to global climate forcing on sedimentation and rock uplift, and three main sequences have been recognized.
 1. The first sequence (PS1; ∼1.4–0.87 Ma) is characterized by meandering river systems prograding eastward and passing downstream to cyclothemic shelfal successions with sedimentation rates of ∼34 cm/kyr outpacing the regional subsidence.
 2. The second sequence (PS2; ∼0.87–0.45 Ma), heralded by a regional unconformity (R surface) correlated to the onset of the major Pleistocene glaciations in the Alps at MIS 22 (∼0.87 Ma), documents a generalized transition to distal braidplain deposition in a low subsidence setting, with marine/continental cyclothems confined to the east of Lake Garda.
 3. The third sequence (PS3; ∼0.45 Ma–present), marked at its base by another regional unconformity, the Y surface, tentatively dated at ∼0.45 Ma using nannofossil biostratigraphy, displays a more proximal trend of braided fluvial deposits, correlative to morainic complexes in the Southern Alps and seemingly deposited under combined conditions of confinement, erosion and bypass. Such conditions are interpreted as produced by the Middle Pleistocene uplift of the Alps.
 4. Excluding local, short-wavelength deformations due to active thrusts, the regional uplift and loss of accommodation space recorded by PS3 along the proximal Po basin are interpreted as the effect of the flexural unloading of the eroding Alpine orogen, with a wavelength of ∼270 km and estimated erosion values ranging from 1.3 to 1.7 mm in the axial sector to 0.1–0.3 mm at the Alpine margins.
 5. As a consequence of uplifting during the Middle Pleistocene, the northern Po basin became a mainly bypass area for sediments delivered by the Alps. As a novel conclusion, the Y surface is interpreted to mark the switch from a depositional stage to an uplift stage at ∼0.45 Ma, an interpretation that is not substantially altered even allowing an older age for the Y surface of ∼0.63 Ma [Garzanti et al., 2011].
 The Middle Pleistocene uplift of the proximal Po basin occurred possibly up to 0.4 Myr after the intensification of erosion in the Alpine axial chain, which is likely to have started as a consequence of the onset of major Pleistocene glaciations in the Alps at ∼0.87 Ma [Muttoni et al., 2003; Häuselmann et al., 2007]. Evidence of uplift in the Alpine proximal peripheral basins after 0.78 Ma can be recognized in southeastern France [Dubar and Semah, 1986] and in Austria [van Husen, 2000], suggesting that different tectonic domains experienced the same event. Climate-driven uplift in the Alps is registered possibly also in the Swiss foreland, where a phase of pronounced fluvial incision, termed Middle Pleistocene reorganization, was observed and tentatively constrained around the Early–Middle Pleistocene boundary [Preusser et al., 2011].
Appendix A:: Regional Seismic Data Set
 The data set used to investigate the geometry and the regional extent of the R and Y surfaces is described in detail by Di Dio  and Carcano and Piccin . It consists of ∼40,000 km of reflection seismics and stratigraphies from ∼330 wells for hydrocarbon exploration, made kindly available by Eni E&P. This data set was updated with 10 cores drilled by Regione Lombardia and discussed by Muttoni et al. , Scardia et al. , and this study. The seismic stratigraphy was interpreted with the aid of Landmark software packages. The lithostratigraphy of the studied Eni E&P wells was deduced by cuttings and borehole logging, including Electrical Survey (ES), Induction Electrical Survey (IES), Spontaneous Potential (SP), and sonic logging.
Appendix B:: BSC Line Data Acquisition and Processing
 Data acquisition was carried out with a fixed cable configuration consisting of 5 m geophone spacing in the northern half of the line and 10 m spacing in the southern half, respectively. The data were acquired with a sampling rate of 1000 Hz, a recording window of 1.024 s, and acquisition groups made up by a double string of 14 Hz geophones. The source was a 120 kg accelerated mass with hydraulic control (Minipulse), 2.5 m spacing in the northern part of the line and 5 m in the southern part. The reflection data were processed by PROMAX and Visual_SUNT software packages. The velocity analysis was performed along the line using the semblance method every 25 common depth points (CDP) (about 75 m). The sorted CDP were corrected of normal move out (stretch mute factor of 1.5) using the velocity profiles obtained in the velocity analysis step. In order to control the near surface seismic velocities along the line (uppermost 60 m of depth), the picked first arrivals were inverted applying the SeisOpt Pro inversion code [Pullammanappallil and Louie, 1994] obtaining a 2-D velocity model.
Appendix C:: Paleomagnetic Analyses
 Laboratory analyses were conducted at the Alpine Laboratory of Paleomagnetism (Peveragno, Italy). Samples from cores RL8 and RL10 were subjected to rock magnetic analysis by means of IRM acquisition curves up to 2.5 T and thermal demagnetization of a three component IRM [Lowrie, 1990] along sample orthogonal axes in 2.5 T, 0.4 T, and 0.12 T fields. The natural magnetic remanence (NRM) was measured on a 2G-Enterprises DC-SQUID cryogenic magnetometer located in a magnetically shielded room. Thermal demagnetization was carried out with an ASC TD48 oven by adopting a minimum of 12 steps from 100 to 600°C up to a (rare) maximum of 680°C. In core RL10, specimens were also demagnetized by alternating field (AF) treatment by 2G-Enterprises Degausser to test the consistency of the paleomagnetic directions obtained with thermal demagnetization. Standard least squares analysis [Kirschvink, 1980] was used to calculate component directions from selected segments of thermal demagnetization diagrams [Zijderveld, 1967].
 This study has been funded by Regione Lombardia in the framework of the CARG project, Sheet 118 Milano. The authors wish to thank Eni E&P for providing subsurface data and the Swiss Federal Office of Topography for the geodetic data. S. Rosselli and L. Papani helped collect and organize core data. S. Barba, G. Monegato, V. Picotti, I. Rigamonti, A. Schlatter, and L. Trombino are acknowledged for useful comments and discussions. We thank the Editor O. Oncken, J.-D. Champagnac, and an anonymous reviewer for critical comments to the manuscript.